An arm and a leg

When SoftBank bought Arm in 2016 for £24bn, the chip designer was delisted from the London Stock Exchange the next day.^{3Silicon RepublicSoftBank finally completes £24bn ARM takeover5 Sept 2016} On Thursday it came back, in New York. Arm priced 95.5m American depositary shares at $51 each and began trading on the Nasdaq under the ticker ARM, raising about $4.9bn.^{1ArmArm Announces Pricing of Initial Public Offering13 Sept 2023,2SEC EDGARArm Holdings Ltd, final prospectus (Form 424B4)14 Sept 2023} It is the sort of homecoming in which the home has changed address.

The case for the shares is easy to state. Arm designs the instruction sets and processor cores that sit inside nearly every phone: its central processors power “greater than 99%” of the world’s smartphones, according to the prospectus.^{2SEC EDGARArm Holdings Ltd, final prospectus (Form 424B4)14 Sept 2023} It does not make chips. It licenses the design, collects a royalty each time a customer ships a chip that uses it, and spends little on anything heavier than engineers. The business has been supplying the industry for more than 20 years to its ten largest royalty customers, on average.^{2SEC EDGARArm Holdings Ltd, final prospectus (Form 424B4)14 Sept 2023} That is as close to a toll road as the chip trade offers.

The numbers are less cheerful than the pitch. In the year to 31 March Arm’s revenue was $2.68bn, against $2.70bn the year before, which the filing itself calls “relatively flat”. It made net income of $524m.^{2SEC EDGARArm Holdings Ltd, final prospectus (Form 424B4)14 Sept 2023} In the quarter to 30 June revenue fell to $675m from $692m, and net income to $105m from $225m.^{2SEC EDGARArm Holdings Ltd, final prospectus (Form 424B4)14 Sept 2023} A growth story needs growth, and Arm’s story, so far, is mostly a story about the next few years. We think investors who pay up for it are making a bet on artificial intelligence rather than on the accounts.

The prospectus also contains the sort of detail a roadshow tends to leave on the slide before last. Arm China, a joint venture that, in the filing’s words, neither Arm nor SoftBank controls, was its single largest customer and accounted for about 24% of revenue. The top five customers, Arm China included, supplied about 57%.^{2SEC EDGARArm Holdings Ltd, final prospectus (Form 424B4)14 Sept 2023} Smartphones and consumer electronics were more than half of royalty revenue.^{2SEC EDGARArm Holdings Ltd, final prospectus (Form 424B4)14 Sept 2023} The risk factors concede that customers “may choose to utilize” free, open-source architectures such as RISC-V instead.^{2SEC EDGARArm Holdings Ltd, final prospectus (Form 424B4)14 Sept 2023} That is the case for competition made by the company itself, in a document it was legally required to write.

The structure of the sale deserves a look as well. Arm receives none of the proceeds. The seller is a subsidiary of SoftBank, which will still own about 90.6% of the company afterwards.^{1ArmArm Announces Pricing of Initial Public Offering13 Sept 2023,2SEC EDGARArm Holdings Ltd, final prospectus (Form 424B4)14 Sept 2023} Ten investors, among them Apple, Intel, Nvidia, Samsung and TSMC, have indicated interest in buying up to $735m of shares at the offer price.^{2SEC EDGARArm Holdings Ltd, final prospectus (Form 424B4)14 Sept 2023} It is touching to see Arm’s customers hold shares in their supplier. It is also what a toll road looks like when the drivers are asked to chip in.

Nvidia’s presence on that list is a small joke at its own expense. In September 2020 it agreed to buy Arm for $40bn, and in February 2022 gave up, citing “significant regulatory challenges”.^{4NVIDIANVIDIA to Acquire Arm for $40 Billion, Creating World's Premier Computing Company for the Age of AI13 Sept 2020,5NVIDIANVIDIA and SoftBank Group Announce Termination of NVIDIA's Acquisition of Arm Limited7 Feb 2022} SoftBank kept the $1.25bn prepayment Nvidia had made.^{5NVIDIANVIDIA and SoftBank Group Announce Termination of NVIDIA's Acquisition of Arm Limited7 Feb 2022} Competition authorities blocked the sale of a neutral supplier to one of its biggest customers, and we think they were right. A public Arm, answerable to many shareholders and with its customers on the register, is the better outcome. It is also the outcome that regulators, in blocking the sale, did not have to design.

The strongest objection is that all this makes Arm look too tidy a bet. The quality of a monopoly-like franchise is real, and so is a licensing model that earns a royalty on every phone sold. But a 90.6% owner is not a free float, and the price of the stock must reflect the people who control it as well as the engineers who built it.

Buyers get less than a tenth of the company. SoftBank keeps the say.